Update app.py

app.py

@@ -41,15 +41,9 @@ Chroma().delete_collection()
 
 for file in os.listdir(path):
     loader = TextLoader(f'{path}{file}',encoding='unicode_escape')
-    # loader.load()[0].metadata['category']='biography'
-    # print(loader.load()[0].metadata)
     documents += loader.load()
-# print(documents)
-print(len(documents))
-'''This is the code used for without memory chat'''
 text_splitter = CharacterTextSplitter(chunk_size=500, chunk_overlap=0)
 texts = text_splitter.split_documents(documents)
-
 embeddings = CohereEmbeddings(model='embed-english-v3.0')
 docsearch = Chroma.from_documents(texts, embeddings)
 retriever=docsearch.as_retriever(search_kwargs={'k':1})
@@ -64,8 +58,6 @@ compression_retriever = ContextualCompressionRetriever(
 memory=ConversationSummaryMemory(
     llm=cohereLLM, memory_key="chat_history", return_messages=True
 )
-# qa = ConversationalRetrievalChain.from_llm(llm=cohereLLM, chain_type="stuff", \
-#     retriever=docsearch.as_retriever(search_kwargs={'k':1}),return_source_documents=True)
 question_generator = LLMChain(llm=cohereLLM, prompt=CONDENSE_QUESTION_PROMPT)
 doc_chain = load_qa_with_sources_chain(cohereLLM, chain_type="refine")
 
@@ -77,9 +69,6 @@ rag_chain=chain = ConversationalRetrievalChain(
 )
 #
 
-# qa = RetrievalQA.from_chain_type(llm=cohereLLM, chain_type="stuff", \
-#     retriever=docsearch.as_retriever(search_kwargs={'k':1}),return_source_documents=True)
-
 btuTuples=pickle.load(open('./bookTitleUrlTuples.pkl','rb'))
 bookTitleUrlDict={x:y for x,y in btuTuples}
 chat_history = []
@@ -91,36 +80,10 @@ def predict(message, history):
     
     message="you are a language model that gives book recommendation based on your context"+message+\
     'just give the book title and author'
-    # history_langchain_format = []
-    # for human, ai in history:
-    #     history_langchain_format.append(HumanMessage(content=human))
-    #     history_langchain_format.append(AIMessage(content=ai))
     result=ai_msg = rag_chain({"question": message, "chat_history": chat_history})
     chat_history.extend([HumanMessage(content=message), AIMessage(content=ai_msg['answer'])])
-    # history_langchain_format.append(HumanMessage(content=message))
-    # result=gpt_response = qa({'query':history_langchain_format})
-    # return gpt_response['result']
-    # gpt_response = qa({'query':''.join(history)+f'.\n given the previous conversation respond using the following prompt:{message}'})
-    # # print(gpt_response)
-    # history.append((f'HumanMessage:{message}',f'AIMessage: {gpt_response},'))
-    # # history=history_langchain_format
-    # return gpt_response['result']
-    
-    # result = qa({"query": message}) # uncomment this to get back to original state
-    
-    # r1=docsearch.similarity_search_with_score(query=q,k=3)
-    # print([(item[-2].metadata,item[-1]) for item in r1],\
-    #       '\n\n',result['result'],f'|| {result["source_documents"][0].metadata}','\n*****\n')
-    # if result['answer'] not in ["I don't know","I don't know."]: #change key from 'answer' tor 'result'
-    #     bookNamePath=result["source_documents"][0].metadata["source"]
-    #     return result['answer']+f'\n---\n***Ignore the description below if the chatbot was unsure about its response \
-    #     or if the response is not about the book shown below. Note that this is a zero memory chatbot as I found it performs better this way.\
-    #     Refer to github.com/mehrdad-es/Amazon-But-Better for more info.\n***\nAmazon Kindle ebook description is:\n {result["source_documents"][0].page_content}'+\
-    #     f'\n**from this file:** {bookNamePath}\n'+\
-    #     f'**link==>** {bookTitleUrlDict[bookNamePath.split("/")[-1][:-4]]}'
-    # else:
-    #     return result['answer']
-    return result['answer']
+    bookNamePath=result["source_documents"][0].metadata["source"]
+    return result['answer'] +f'''---\nlink: {bookTitleUrlDict[bookNamePath.split("/")[-1][:-4]]}'''
 
 gr.ChatInterface(predict,
     chatbot=gr.Chatbot(height='auto'),